University of California, Merced
School of Engineering
Spatial Analysis and Modeling (ENGR 180)

Key assignment steps are in RED

Case Study Scenario

Lab 3-1 and 3-2 will introduce different concepts in remote sensing and explore some of the kinds of spatial analysis that can be done with remotely sensed data. You will use satellite imagery downloaded from USGS to explore different methods of evaluating plant productivity. This will culminate in the use of satellite imagery to identify regions of widespread vegetation death due to wildfire.

Be sure to save throughout your process, as you will use 3-1 outputs and project for 3-2 next week.

Introduction

Please read all of the instructions/narrative carefully! You may feel compelled to skim to find which elements you have to satisfy to get full credit. If you skim, you may miss key sentences and you may end up doing more work than necessary! Also, if you have questions, give the Case Study handout a second reading. Your question is probably answered here!

Data Management WARNING: Raster datasets are often large! Depending on how they were encoded, they may also compress well. You will need to extract all .zip files in order for the files to be read by ArcMap, or Excel (or most any other program). Be mindful of the available disk space on personal devices, because some rasters expand to hundreds of megabytes when unzipped. You will need a lot (at least 3GB) of free space for this lab. Have a plan for managing your data in your labs.

If text is in italics, it’s a tool or step within ArcGIS Pro.
If text is underlined, it’s a file.

Part 1: Sources of satellite imagers

We are using imagery from Landsat 8. Please read Section 1 – Introduction from the “Landsat 8 (L8) Data Users Handbook” [1] for a background on the Landsat program.

Satellite imagery can be downloaded from a variety of sources including:
• USGS GloVis (Terra-ASTER, EO-1-{ALI, Hyperion}, RESOURCESAT-{AWiFS, LISS-3}, Landsat 1-5 7-8, OrbView-3, Sentinel-2) https://glovis.usgs.gov/

• USGS Earth Explorer (a lot of imagery, including all AVHRR, IKONOS-2, OrbView-3, Declassified Keyhole Satellite data, ASTER/EDNA/GTOPO30/and LIDAR DEMs, National Atlas maps, EO-1, Landsat data, Sentinel-2 data, RESOURCESAT data and more) https://earthexplorer.usgs.gov/

• The Copernicus Open Access Hub (Sentinel-1, Sentinel-2, and Sentinel-3)

• NASA Earth Data Search (formerly Reverb data search| ECHO, houses current and historical NASA Earth Observation Satellite data) https://search.earthdata.nasa.gov/

For this lab, you are provided with Landsat 8 scenes from July and October 2017 taken under clear conditions, with the sun nearly completely overhead.These images were collected using a free account through EarthExplorer, a USGS product by searching the date, region of interest, and desired cloud cover for Landsat Level 2, Collection 2.

Part 1b: Data Download

Download your data from the following Box Link. These are large files and may take a few minutes. This would be a great time to prep a word doc for your lab responses!

These zipped folders contain multiple individual files - click into the files, and you’ll notice names like LC08_L2SP_043034_20171013_20200902_02_T1_SR_B3.

B3 refers to Band 3, which is visible green (anything else that doesn’t appear to our eyes as green was filtered out by the sensors). Per Landsat, this is the full naming convention:
“The Landsat Collection 1 Level-1 product identifier includes the Collection processing levels, processing date, collection number, and collection tier category:

LXSS_LLLL_PPPRRR_YYYYMMDD_yyyymmdd_CC_TX

Where:

    L = Landsat
    X = Sensor (“C”=OLI/TIRS combined, “O”=OLI-only, “T”=TIRS-only, “E”=ETM+, “T”=“TM, “M”=MSS)
    SS = Satellite (”07”=Landsat 7, “08”=Landsat 8)
    LLL = Processing correction level (L1TP/L1GT/L1GS)
    PPP = WRS path
    RRR = WRS row
    YYYYMMDD = Acquisition year, month, day
    yyyymmdd - Processing year, month, day
    CC = Collection number (01, 02, …)
    TX = Collection category (“RT”=Real-Time, “T1”=Tier 1, “T2”=Tier 2)”

1. Download the full folder https://ucmerced.box.com/s/2a50ax568nkza4mpz6imfwlsbjfe8swc to your workspace folder. Be sure to unzip the folder!
2. Rename the downloaded folder to LandsatImagery

Though the file names within the folder look very similar, note the date in the middle to differentiate between the two datasets.

• July, 2017 Imagery: LC08_L2SP_043034_20170709_20200903_02_T1
• October, 2017 Imagery: LC08_L2SP_043034_20171013_20200902_02_T1

Part 1c: Types of satellite imagery

For multispectral satellites, different bands correspond to different regions of the electromagnetic spectrum. The pixel value (roughly) corresponds to an average brightness within a particular range of the electromagnetic spectrum. For a visual understanding, refer to the Fig. 1 [2].

Figure 1: Electromagnetic spectrum: Bands and Satellites

The Landsat scenes that you will be using are Level-2 products, where the digital value (pixel values) have been transformed into surface reflectance after correcting for atmospheric scattering, solar angle and other sources of bias.

Surface reflectance represents the fraction of total incoming solar radiation that makes its way back to the Landsat sensor. For more on Level-2 (derived) Landsat products, refer to the Landsat Surface Reflectance Level-2 Science Data Products article [3]. The band designations, wavelengths, and resolution information for Landsat 8 are shown in Table 1.

Table 1: Landsat 8 designations, wavelengths, and resolution information

Part 1d: Band symbology

Common digital cameras are sensitive across a range of wavelengths, extending from the blue region of the electromagnetic spectrum (EM), down to Near-Infrared (NIR).

A Bayer filter in the camera filters out different wavelengths of light so that a given pixel is sensitive to a particular band (red, green, or blue). The digital camera superimposes the different bands to produce a color, RGB image. You can also produce a color image from satellite products manually in ArcGIS Pro.

Part 2: Band visualization

Part 2a: Opening images in Pro

  1. Open ArcGIS Pro and begin a new project for Lab 3-1.
  2. In Catalog, connect to your workspace by right clicking on Folders and Add Folder Connect, then select the folder containing the downloaded Landsat data.
  3. Right click on Folders in the Catalog Pane and Add Folder Connection
  4. Connect to your LandsatImagery folder
  5. Open the Red, Green, and Blue bands of the July imagery by dragging and dropping the corresponding bands into the Map window: (Bands 4, 3, and 2: the order in which you bring them in matters!)
  6. If asked, click Yes to Build Pyramids (and/or Calculate Statistics). For more on the purpose of raster pyramids, refer to the Raster pyramids article in the ArcGIS documentation [5].

Part 2b: Produce componsite image

We can combine individual bands to create a full color composite (i.e. “normal” looking photo)

  1. Access the Composite Bands tool through The Analysis Ribbon by searching the tool box.
    • Note: “The order that the bands are listed in the Multi-value Input control box will determine the order of the bands in the output raster dataset.” This is important. Input the bands in the R-G-B order.

  2. Name the output Comp_432 (See Fig. 2)

  3. Set your Input Rasters to contain Bands 4 (R), 3 (G), 2 (B)

  4. Run

Figure 2: Composite image output


Create and export a production quality map of your output, with the location of UC Merced visible and labeled (it can be as simple as inserting a text box next to the location of UC Merced). Remember, a screenshot of an image is not a map.

Part 3: Map algebra - band math

There are many spectral indices that are useful in their ability to be correlated with different natural features of index. Many of these indices are expressed as a ratio, where one band is divided by another. Normalized difference indices are often represented as the difference between two bands, divided by the sum of the bands, producing a ratio image where majority of values lie between -1 and +1. These include the following:

• Normalized Difference Vegetation Index (NDVI), an index of relative photosynthetic activity, often used to highlight vegetated areas from the rest of the landscape or to evaluate plant productivity.

• Normalized Difference Water Index (NDWI), an index that is correlated with water content in water bodies.

• Normalized Difference Moisture Index (NDMI), an index that is associated with vegetation moisture. It can be used in drought monitoring.

• Normalized Burn Ratio (NBR), an index that highlights differences between live and burned vegetation.

In this section, we will use the Raster Calculator to compute different vegetation indices in the Merced region.

Part 3a: Spectral indices - NDVI

The Normalized Difference Vegetation Index (NDVI) is calculated as follows

_NDVI Equation
where NIR = Band 5 and Red = Band 4 on the Landsat 8 Operational Land Imager (OLI).

Use the Raster Calculator to produce an NDVI raster for the July 2017 Landsat scene

  1. Bring in Raster Bands 5 and 4 for the July Landsat data (like step 3 in Section 2a)
  2. Wrap the numerator and denominator (each) in the Float function, to produce a floating point (read: fractional) output raster
  3. Name your output tm_1707_ndvi
  4. Use a “Green-Red” color ramp, where red is associated with negative values and green with positive values (see Fig. 3).
    • Super hint: Your function will look something like this:

NDVI Calculation

Figure 3: NDVI Results

Create and export a production quality map of your result. Describe what positive and negative values of NDVI are correlated with (refer to the NDVI function documentation [8]. Be sure to properly cite any sources!

Part 3b: Spectral indices- NBR

The Normalized Burn Ratio (NBR) is calculated as follows: NBR Equation

where SWIR = Band 7 and NIR = Band 5 on the Landsat 8 Operational Land Imager (OLI).

  1. Use the raster calculator to produce an NBR raster for the July 2017 Landsat scene
  2. Again, wrap the numerator and dominator respectively in the Float function
  3. Name your output tm_1707_nbr
  4. Symbolize your map with a contrasting color ramp, such as “Pink to YellowGreen”, where green corresponds to positive values and red/pink corresponds to negative values.

    Create and export a production quality map of your result. Describe what positive and negative values of NBR are correlated with (refer to the HSU article on Normalized Burn Ratio linked in the references [9])

Part 3c: Spectral indices- Change over time

We are going to work with a lot more layers in this assignment - we can “compartmentalize” our data in the Contents pane by right clicking on Map in Contents, then creating a New Group Layer.

  1. Drag and drop your July files into a Group Layer named July.
  2. Bring in all relevant October bands into a Group Layer named October.
  3. Compute the NBR for the October 2017 Landsat scene.
  4. Name your output tm_1710_nbr
  5. Compute the difference between the July and October NBR rasters with Raster Calculator
  6. The difference (dNBR) is computed as follows:

Δ𝑁𝐵𝑅=𝑃𝑟𝑒𝑓𝑖𝑟𝑒 𝑁𝐵𝑅−𝑃𝑜𝑠𝑡𝑓𝑖𝑟𝑒 𝑁𝐵𝑅 (Eq. 3)

  1. Name your output tm_Dnbr

The Detwiler Fire burn region (between Lake McClure and Mariposa, CA) should be very apparent in the October NBR and in the difference NBR

  1. Reclassify your difference raster according to Table 2.To do so, open the Reclassify (Spatial Analyst) tool in the Analysis Ribbon.

9.Click the Classify button beneath the empty table. Set the classification to manual.

Table 2: Classification Ranges

Name your output tm_Dnbr_cls

  1. Symbolize your map with a contrasting color ramp with redder/more dramatic tones to the right, such as “Condition Number”, as seen in Fig. 4.

  2. Import the Detwiler Fire boundary into your project geodatabase. This file can be found in the Lab data folder in CatCourses, under: ca_detwiler_20170731_0600_dd83.zip

    Create and export a production quality map of your NBR difference, focusing on the Detwiler Fire region.
    In writing, compare your raster to the Detwiler Fire burn perimeter - zoom in very closely if needed.

Figure 4: PreFire - PostFire NBR

Part 4: Spatial Analysis for problem solving

Mariposa County hired you to revise their regional evaluation of burn risk after the Detwiler Fire. You decide to see if burn severity is related to the following two categories:
• Landscape factors, such as vegetation health (from NDVI)
• Human factors, such as distance from roads (from a road network)

You decide to use the Zonal Statistics tool to generate summary statics within the different burn severity zones. Your final report to the local commissioner will include tables which describe the mean, median, and maximum:
• Pre-fire NDVI within each burn severity zone
• Euclidean distance from roads within each burn severity zone

Part 4b: Pre-Fire NDVI

You have already calculated pre-fire NDVI in Part 3a, with the output saved as tm_1707_ndvi.

Part 4c: Euclidean distance from Roads

Use the Euclidean Distance tool [11, 12] to produce a raster that represents Euclidean distance from roads. Use the following workflow:

  1. Import the 2015 TIGER/Line Shapefile for California, Primary and Secondary Roads into your project geodatabase. Once it’s imported, it’s called a feature class.
    • The shapefile is found under: tl_2015_06_prisecroads.zip

  2. Reproject the feature class to match the CRS of your Landsat scene (refer to Lab 2-1 for a refresher on the “Project (Data Management)” tool. Name this roadsPRJ

  3. Clip roadsPRJ to the extent of your Landsat 8 scene. Name this clippedRoads
    • Hint: Use the “Raster Domain (3D Analyst)” to create a polygon that represents the extent of your unprocessed/original Landsat 8 scene.

  4. Rasterize the road feature class using “Feature to Raster (Conversion)” tool. Name your output roadsRAS
    • Hint: Use the RTTYP (road type) as your attribute Field
    • Cell size should match the cell size of your Landsat scene (30 meters)

  5. Use the “Euclidean Distance (Spatial Analyst)” tool, to generate a raster that represents distances from your roadsRAS road network
    • Hint: The only input will be your rasterized road network. You do not need to specify any other input parameters for this tool

  6. Save your raster as euc_road_ras. This may take a minute or two.

  7. Knowing we are going to do statistics on this layer in the next section, we have one final step. Our pixels in euc_road_ras are floating point, but the upcoming statistics tool only accepts integers. Use Int (3D Analyst Tools) to convert your euclidean roads distance pixels into whole numbers. Save the output as euc_road_rasINT

Export an image of your Euclidean distance raster with your clipped road network overlaid (see Fig. 5). Screenshots are OK here - you do not need a full map.

Figure 5: Euclidean road distances

Part 4d: Zonal Statistics

Use the “Zonal Statistics as Table (Spatial Analyst)” [13, 14] to summarize the NDVI and distance rasters, by burn category.

Prep: Import ca_detwiler_20170731_0600_dd83 into your Geodatabase and name it DetwilerBounds.

  1. This is very important. Clip your categorized NBR raster (tm_Dnbr_cls) to the burn perimeter polygon DetwilerBounds using “Clip (Data Management)”. Make sure to check “Use Input Features for Clipping Geometry” in the Clip tool. If you do not clip it, it will compute statistics for the entire Landsat tile VERY SLOWLY.
    • Hint: Your input raster should be the categorized NBR raster
    • Hint: Your input value raster should be the NDVI or distance rasters
  2. You will have to run the tool 2 times
    • Calculate all summary statistics: Name your outputs NBRNDVIZS and NBREUCZS.

Present your summary statistics in a tabular form and discuss your findings. Hint: Try Table to Excel to export the statistics. What do the tables tell us?

Prepare a production quality map of the NBR zones overlaid on the Landsat 8 color imagery (or a background of your choosing)

Part 5: Deliverables

At the end of this Case Study, you should have generated the following:

- Part 2b: Production quality map of Composite image with UCM visible and labelled
-Part 3a: Production quality map of July NDVI with accompanying description of what positive and negative NDVI values correlate with.
-Part 3b: Production quality map of July NBR with accompanying description of what positive and negative NBR values correlate with. Be sure to properly cite any sources!
-Part 3c: Production quality map of NBR difference in Detwiler Fire region. Written comparison of your results to the burn perimeter results.
-Part 4c: Screenshot of your Euclidean distances from roads
-Part 4d: Table of your summary statistics and production quality map of NBR zones overlaid on color (composite) Landsat image or basemap of your choice.